RE: No Standard Semantic Web Pragmatics?

> From: Pat Hayes
> Sent: Friday, June 18, 2004 3:07 PM
> 
> >At 13:41 -0400 6/17/04, Lynn, James (Software Services) wrote:
> >>Jim,
> >>
> >>Just to make sure I understand your position regarding this, can 
> >>you tell me or give an example of how I would know, or even be 
> >>relatively sure, whether someone is making an observation about a 
> >>statement or accepting that statment in what some might call a 
> >>legally binding manner. The statement "I agree with these terms and 
> >>conditions" comes to mind, but use your own example if you wish? 
> >>Does this statement enter into a legal contract? Is this simply a 
> >>matter of using the correct definition of the word "agree" as in 
> >>legally_binding:agree vs. myopinion: agree?
> >>
> >>Thanks,
> >>
> >>James Lynn
> >
> >
> >James - there are times when saying that does indeed cause me to 
> >enter a contract, but there's times it doesn't -- and the words 
> >themselves (separate from the social context they appear in) can 
> >never tell me which it is.
> 
> No, they can: we humans DO use words to do just this. They work 
> because of some kind of ultimate social contract, if you like, but in 
> many cases this is no more than the contract of using the conventions 
> of ones native language. If I say "I promise..." then I do, indeed, 
> thereby promise: I perform a social act by using a word in a certain 
> way. And this is part - a large part  - of what the word "promise" 
> MEANS in English: it can be used to make promises, as well as to talk 
> about - describe - promises and promising. So given this basic level 
> of 'social context' one can do things like enter into contracts just 
> by saying one is entering into it.  So...
> 
> >So I just don't believe we can create a "syntactic" way to tell 
> >these things apart
> 
> .... we CAN create such a 'syntactic' way. In fact, we have done: the 
> details are in the paper.
> 
> >-- given that, what is a different choice?  It's kind of a "client 
> >decide" mode
> 
> But it can't be 'client decide' if there is no way for the server to 
> tell the client. At best it will be 'client guess'.

It's both. Your both right. URI references and imports are like Janus,
http://concise.britannica.com/ebc/article?eu=393749
both client AND server. They have two faces, one face looks back as a 
client at existing work, the other face looks forward as a server of a 
new work. They are the doorways back to those whose thought we build 
upon and the facade upon which we present our own thought. No wonder 
it is hard to get it right. In my opinion, there can be no simple, one
size fits all convention for how meaning* should pass from author to 
author through the gateways of URI references. Each author needs more 
control over that flow* than a simple on and off valve. If you must, 
when you open the valve of a URI ref, let everything enter your own 
document, people will refuse to open that valve. But if you shut it 
off completely, you end up secluded in self-protected solipsism. 

In this exchange, even though he calls it 'client decide' Jim seems to be 
thinking of what the client will do, as a server, with the stuff it reads, 
e.g. search agent, etc. Pat points out, apparently thinking of how the 
client is reading stuff with no intentionality expressed, how the client 
can do no more than guess what that stuff should count as. What I'm 
trying to say is that, in the present specs, what we have is 'client 
guess blindly AND client decide mutely'. There is no way for the original 
publisher to tell us how to USE some RDF, nor for any Janus 
reader/publisher to tell us how it wants to REUSE that original stuff to
make new stuff.

We need to give authors a voice so they can act as the gatekeeper of 
their URI REFERENCES. We need to give them URI SPEECH ACTS. So get out 
and vote for Standard Semantic Web Pragmatics.

John Black

*intentionally metaphorical use of the word 'meaning'
 
> >-- i.e. a search engine may not care, a legal aid agent will try to 
> >determine, a contracting agent will insist that some extra stuff, 
> >outside the simple context (i.e. something more than "is it on the 
> >page" -- maybe something like "is it on the page in a form that 
> >matches a previous agreement of how such legally binding pages will 
> >be represented and containing a trusted key signed by one of the 
> >people i have previously negotiated my online contracting with or 
> >authorized to serve as an authorization proxy for me"
> 
> The stuff in the paper is pretty much exactly a proposal for doing 
> this in a machine-recognizable way. That is all we are talking about.
> 
> >)  Strangely, experience would seem to show that in some ways these 
> >more complex things are actually "simpler" in a certain sense.  That 
> >is, the phrase "I do" doesn't commit you to a legal contract outside 
> >of the context of being in response to a certain question asked by 
> >someone who has been certified by a legal authority (usualy the 
> >State) to represent them in this kind of contracting - which makes 
> >it a lot easier to tell a real marriage from a couple of kids 
> >fooling around on the sandlot -- okay, so what has this got to do 
> >with the Sem Web?  Well, I guess I would like to see a situation 
> >something like "If I produce a semantic web document (or any variant 
> >thereof that is parsable), and place it on the Semantic Web,  you 
> >are entitled to believe I has created those triples on the Semantic 
> >Web, and nothing more."
> 
> But this is unrealistic. Try telling that idea to the WebServices 
> group, for example.  With this reading, no agent would be justified 
> in doing anything with any SW content. This reduces the SWeb to a 
> bunch of meaningless stuff for academics to play with.
> 

And if you insist on this type of self-protective isolation, leave it 
as the default, but add optional facilities for those of us who want 
to do more, and include a standard explanation of How to Do Things 
With Triples in the specification. 
http://www.amazon.com/exec/obidos/tg/detail/-/0674411528/104-2926588-0135148?v=glance

> >If, based on some other, non SW context, or some prearranged or 
> >authorized other context there is an expected social agreement, than 
> >I should make that clear in a human readable form,
> 
> But why do you insist that it be a non-SW context? And why should it 
> have to be  not-machine-readable?
> 
> >or in a form for which some body with "social clout" has determined 
> >should have a greater meeting (for example a Creative Commons 
> >license).   Thus, I'd argue that a Creative Commons license becomes 
> >meaningful by dint of being a CC license (and there being a 
> >recongizable way to tell), not by dint of being some random triples 
> >in RDF.
> >  Now, all that said, I would argue that some minimal levels of 
> >agreements could produce the necessary 80% needed for this thing to 
> >grow and thrive (Since very few sites will need the RDF to be 
> >"legally binding" in any sense).  So, for example, I would like to 
> >see something that says (better than this) that if I create a 
> >document that uses a relation defined on another document (and named 
> >via a URI), then it is expected that the named document is 
> >dereferencible if the document uses RDFS or OWL terminology  and, I 
> >would probably be willing to consider the things implied by that 
> >document (under the RDFS and OWL semantics) to be the same as above 
> >(i.e. that you are entitled to believe that I am pointing at a 
> >document containing those triples and nothing more barring social 
> >agreement)
> >  So, what do I think is our job?  Identifying some of these minimal 
> >social conventions that will help make this work better.  What do I 
> >want us to avoid?   Arguing philosophical issues about "meaning" and 
> >"intent" which have no impact and don't bring us forward in any way
> 
> The only one here arguing about philosophy is you. What John wanted, 
> and what Jeremy and Patrick and Chris and myself are proposing, is a 
> way to embody these minimal social conventions in workable 
> engineering using existing standards and conventions.  It is to some 
> extent informed by linguistic philosophical thinking, but I find that 
> encouraging: it suggests to me that we may have succeeded in formally 
> capturing a bit of the way that language operates 'naturally'. It 
> felt like a step forward to me. I don't know why the hell you are 
> giving us all such a hard time about it.
> 
> Pat
> -- 
> ---------------------------------------------------------------------
> IHMC	(850)434 8903 or (650)494 3973   home
> 40 South Alcaniz St.	(850)202 4416   office
> Pensacola			(850)202 4440   fax
> FL 32501			(850)291 0667    cell
> phayes@ihmc.us       http://www.ihmc.us/users/phayes
> 
>

Received on Saturday, 19 June 2004 08:24:46 UTC